Andreessen Horowitz has raised $1.1 billion for a new vehicle called the Machine Age Fund, the firm announced on August 28, 2026, dedicating the capital to the physical infrastructure that artificial intelligence runs on.
The fund was announced in a post on the firm’s website authored by general partners Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George. The partners framed the raise as a formal move into hardware investing for a firm generally known as a software investor, writing that “it’s time to open the throttle and accelerate the physical buildout of AI.”
What the Fund Will Invest In
According to the announcement, the Machine Age Fund will invest in the computer infrastructure on which AI runs, naming chips, memory, networking, and storage as target categories. The mandate also covers full systems for running AI, spanning data centers, robotics, and home AI appliances.
The partners wrote that every layer of this stack is hitting the limits of today’s supply chain capability and of physics and computer science. They described the fund’s scope as extending down to the electrical level, calling the current moment a generational opportunity to rearchitect these layers as platforms.
The Infrastructure Constraints a16z Cites
The announcement laid out specific figures describing how infrastructure demands have escalated. The partners wrote that compute density per rack increased 28-fold from an H100 rack to a Rubin rack, and that networking within a rack has grown to the point of reaching the limits of copper cabling. Rack power has moved from roughly 5 to 10 kilowatts to 100 to 250 kilowatts for current systems, the post stated, and is expected to reach 1 megawatt within three years. Data center scale is moving from tens of megawatts to hundreds of megawatts and, in some cases, to gigawatt-scale campuses, with power increasingly coming from behind-the-meter or captive sources alongside the grid.
The linked figure on rack power points to NVIDIA’s 800 VDC architecture documentation, in which NVIDIA describes a power-delivery redesign intended to support 1-megawatt IT racks starting in 2027. That NVIDIA post notes that today’s 54-volt in-rack power distribution, designed for kilowatt-scale racks, is not suited to megawatt-scale systems, and that a single 1-megawatt rack using 54-volt distribution would require up to 200 kilograms of copper busbar.
The a16z partners wrote that the industry needs faster and more efficient systems, cheaper and higher-bandwidth memory, faster and more scalable interconnects, and power-efficient edge devices, along with the cooling, materials, electrical, and real estate buildout to support them. They stated that the hardware supply side is accustomed to growing 20% to 30% per year at most, rather than the triple-digit growth they said is now required to catch up with demand.
Hardware Deal Flow and a16z’s History in the Sector
The announcement said hardware startups have grown from a small share of the firm’s deal flow to more than 20% over the past couple of years. The post named several hardware companies a16z has recently backed, including Unconventional AI, Nexthop, Volta, Atoms, Heron Power, and Mind Robotics.
The partners also pointed to a longer history of hardware investments at the firm: a16z led the Series A in drone maker Skydio in 2016, invested in SpaceX, wrote its first check into defense contractor Anduril in 2019, and was among the first venture investors in Waymo’s 2020 raise.
On team background, the post noted that Guido Appenzeller was previously CTO of Intel’s Data Center Group, that Raghuram and Casado spent multiple decades in the data center space, that Shangda Xu and David George have led investments across the AI infrastructure stack from silicon and networking to large-scale systems and compute platforms, and that Ulevitch and Erin Price-Wright lead many of the firm’s hardware and U.S. manufacturing investments through its American Dynamism practice.
The firm said its go-to-market, talent, and marketing organization is now prepared to serve hardware founders, and invited founders building in AI hardware to reach out.
Credit: Source link
